Search

In this paper, we propose an extraction method for landslide topographic features and based on these fine features, we trained a landslide stability classification model. Topographic factors have a huge impact on the stability of landslides, and our new feature extraction methods can provide more accurate portraits of landslides. Considering the factors affected landslide stability, this paper studied the factors used in traditional quantitative stability calculations and analyzed the factors affecting stability based on landslide data. So, the proposed method is a combination of theoretical method and numerical method. Based on support vector classification (SVC), random forest (RF), and extreme gradient boosting (XGBoost) models, this paper compares and analyzes the rationality and reliability of the landslide stability classification results output by three models. After adopting the new fine feature extraction method, the accuracy of the stable classification results has been improved by about 5% on average, and the interpretability of the classification results has also been greatly enhanced.

In landslide disaster prevention project, landslide profile is an important research object, which contains rich geological information of landslide. When we get the profile image of the landslide, if we can recommend several landslide cases with similar geological characteristics by the landslide profile image for research, the landslide disaster prevention project can be carried out quickly. In this paper, we propose a recommendation system based on similarity of landslide geological characteristics. The recommendation system is divided into two steps. Firstly, we construct a classification model to obtain the categories of landslide geological characteristics from the profile image by random forest algorithm and supervised machine learning approach. Then, according to the categories of landslide geological characteristics, the landslide case recommended model is constructed, and 5 landslide cases with similar geological characteristics are recommended for the landslide. According to the evaluation index of the recommendation system, the evaluation results show the effectiveness of the recommendation system.

House prices have received unusual attention in recent years, and evaluation of housing price has also become a research hotpot. Value of a house is considered as a mixture of various features, such as house type, orientation and location. Accurate assessment of houses thus faces a large number of challenges. Several methods have been applied to assess the value of a house, including geographically weighted regression and machine learning. However, every method has some limitations. The accuracy of traditional regression methods needs to be improved. Machine learning methods are weak in interpretability. In this paper, to analyze the housing price, we compared XBGoost, Random Forest, Support vector regression, back-propagation network and geographically weighted regression to model the second-hand housing price dataset, and analyzed the influenced of each feature. The result shows Random Forest reaches a lowest mean absolute error of 1517.29 and best r2 of 0.83. According to the constructed models, the importance values are given, and the area of house shows a strongest influence on housing price in the study area.

Electrodynamic suspension (EDS) system based on moving Halbach Array over a conducting plate has been verified to be poor in stability, but it is not clear whether the load could impact on it. This paper focuses on the effect of the load on the stability of the EDS system based on the Halbach array moving over a plate. Generalized energy was introduced by quantifying the Hamiltonian function. By Pade approximation, formulas to obtain both the amplitude and frequency were given. Load varied from 50kg to 64kg, keeping the generalized energy fixed to study the stability. Results show that the stability of electrodynamic suspension system could be improved by lager load, yet the oscillation frequency drifts.

Impervious surfaces have become the most intuitive indicator in the process of urbanization. Timely and accurate information on impervious surfaces from remote sensing images is essential. It not only helps us understand the process of land use/cover change, but also the influences on human society and the environment. In this study, convolutional neural network (CNN) was used to extract the impervious surface in Chengdu city, Sichuan province, China. The overall accuracy in 2009 and 2017 were 98.75% and 99.76% respectively. From the results for 2009 and 2017, the impervious surface increased by 51.24 km2, Growth rate is 13.8%. During the process of urban expansion, suburban farmland was replaced by impervious surfaces and the area of impervious surface gradually increased.

With the rapid development of urbanization, more and more attention has been paid to the structure of urban function zone. Thus, it is of great significance to investigate urban function zone. In this paper, we introduced the deep neural network (DNN) to infer the urban function zone with a supervised classification approach, taking the Shenzhen city in China as a case. First of all, the urban road networks of Shenzhen city were gathered and selected appropriately. Then, the fifth level road networks were utilized to segment the study region. Second, the communication data of different times and points of interest (POI) were collected. Then, the fifteen factors influencing urban function zone were derived. In addition, the urban function zone was divided into five types and the labeled examples with fifteen influencing factors were chosen. Third, the labeled examples were employed to train the DNN with different hidden layers compared with random forest (RF) and support vector machine (SVM). The models were trained with the approach of five-fold cross validation, and the average training accuracy with five times is taken as the accuracy of models. Finally, this paper compared the accuracy. It's been shown in the results that DNN was the optimum model and achieved the highest accuracy. Therefore, our proposed method is an efficient approach to infer the urban function zone.

The number of applications that use radio-frequency identification (RFID) technology has grown continually in the last few years. In this presentation we will present the results of research on the electromagnetic environment on RFID systems. We analyze the effects of UHF MHz China RFID system near a GSM frequency bands. Kurokawa's method of calculating the power reflection coefficient from the Smith chart in the situation is applied to passive RFID tag design. The performance analysis of a specific RFID tag is presented to be compared with the theory. One of our objectives has been to examine tag designs that are less sensitive to their environment, in that their radiation patters and tag response characteristics change little when placed on different distance and orientation. We introduce a technique for deploying RFID systems and engineering solutions based on fundamental electromagnetic understanding. The paper will also explore tag designs that use electromagnetic models to simulate and engineer RFID systems.

Advances in BGA packaging technologies have led to a dramatic increase in the performance of integrated circuits. Noise source such as supply bounce, signal coupling, and reflections results in reduced performance. The work presents techniques to model and improve performance the performance of BGA designs without moving toward advanced packaging. A single, unified mathematical framework is presented that predicts the performance of a given package depending on the package parasitics. Using 3-Bit Bus example about the package, a methodology is presented to analyse mutual inductive signal coupling and mutual capacitive signal coupling and return current. The performance model illustrate that the per-pin performance is significantly reduced, where the number of switching signals is increasing.

Through silicon vias (TSVs) are one of the major enabling technologies for three-dimensional packaging (3DP). TSVs are usually plugged with plated copper for interconnection. However, the introduction of massive copper in the chip may cause device failure due to the fast diffusion of copper into silicon. In order to reduce this risk, a diffusion barrier layer should be deposited on the sidewall of TSVs between the plated copper and the bulk silicon. In addition, an insulation layer should be deposited as well to prevent short circuit. In this paper, the phenomenon of copper-to-silicon diffusion is characterized. The selected interfacial multilayer structure is Cu/Ti/SiO/sub 2/. Although the ideal specimens should be TSVs with interfacial multilayer deposited on the side wall, it is rather difficult to perform surface analysis on a very small area. Therefore, planar analysis with interfacial multilayer deposited on the top surface of a silicon wafer is performed instead. For comparison purpose, additional cases without insulation layer and/or barrier layer are also investigated. A set of experiments is designed to study the effects of surface roughness and layer thickness. In addition, some samples are subject to high temperature storage for the study of thermal annealing effect.

In spatial hearing, complex valued head-related transfer function (HRTF) can be represented as a real valued head-related impulse response (HRIR). Using Karhunen-Loeve expansion, the spatial features of the normalized HRIRs on measurement space can be extracted as spatial character functions. A neural network model based on Von-Mises function is used to approximate the discrete spatial character function of HRIR. As a result, a time-domain binaural model is established and it fits the measured HRIRs well.